Name: Light Concepts
Rating: 5 (117 reviews)
Author: motimalu

physical violence

weapon violence

wide hips

revealing clothes

thick thighs

downblouse

convenient censoring

huge breasts

pg-13

corpses

suggestive

oral invitation

pg13

sexy

sexual situations

male nudity

disturbing

male swimwear or underwear

female swimwear or underwear

partial nudity

undressed

female nudity

breasts out

exposed female nipple

breast out

lingerie

male underwear

hair over breasts

female swimwear

gigantic breasts

no panties

graphic violence or gore

covered nipples

huge butt

strapless leotard

sitting on face

emaciated bodies

one breast out

nsfw

female underwear

nude

graphic male nudity

adult toys

illustrated explicit nudity

nudity

graphic female nudity

hentai

futanari

porn

sexual intent

genitals

peeing

vore

oral

sexual activity

anal

blowjob

dildo riding

incest

hanging

hate symbols

nazi party

white supremacy

diapers

scat

self injury

hate speech

urine

extremist

child on child

latex clothing

swimwear

bukkake

fellatio

cumshot

implied fellatio

eat_cum

cumdrip

cum in pussy

cum on face

after fellatio

cum on hair

cum on body

cum on tongue

cum on hands

cum in mouth

triple fellatio

autofellatio

fucked silly

cum on pussy

Light ConceptsTraining data is a collection of various light concepts I enjoy using that are not overly represented in large datasets, trained as a single lora.<span style="background-color:rgb(26, 27, 30);font-family:-apple-system, BlinkMacSystemFont, &quot;Segoe UI&quot;, Roboto, Helvetica, Arial, sans-serif, &quot;Apple Color Emoji&quot;, &quot;Segoe UI Emoji&quot;;font-size:16px">ℹ️ LoRA work best when applied to the base models on which they are trained. Please read the About This Version <span style="background-color:rgb(26, 27, 30);font-family:-apple-system, BlinkMacSystemFont, &quot;Segoe UI&quot;, Roboto, Helvetica, Arial, sans-serif, &quot;Apple Color Emoji&quot;, &quot;Segoe UI Emoji&quot;;font-size:16px">on the appropriate base models and workflow/training information.I've trained many of these concepts before, generally they are nice to use to enhance the lighting of generations or give interesting effects.<span style="background-color:rgb(26, 27, 30);font-family:-apple-system, BlinkMacSystemFont, &quot;Segoe UI&quot;, Roboto, Helvetica, Arial, sans-serif, &quot;Apple Color Emoji&quot;, &quot;Segoe UI Emoji&quot;;font-size:16px">Trained on a large mixed NL and tags dataset, at mixed [1024, 1536] resolutions. Previews are mostly generated at 1024x1536 with a combination of tags and NL prompts.Concept tagsNot limited to, but collected by works containing:<pre><code>
dispersion
hue shifting
refraction
subsurface scattering
translucent

bioluminescence
caustics
dappled moonlight
glowing hot
ultraviolet light</code></pre>Works best in combination with NL if you name a character, describe their basic appearance, and finish with descriptions of light sources and their effects on the scene:<pre><code>
A vibrant and dynamic illustration of Hoshimachi Suisei from Hololive, featuring her squatting in front of a glowing triangular prism.
A beam of white light enters the prism from the left and refracts into a vibrant rainbow on the right. The background is a solid dark grey to emphasize the lighting effects.</code></pre>

Trained on <a target="_blank" rel="ugc" href="https://civitai.com/models/2458426/anima-official?modelVersionId=2836417">Anima Preview 3 Base</a>With a mix of natural language and tag captions.Training config:<pre><code># trained using diffusion-pipe commit 518ba041d867e6d76c31806a8d0b6a0263fb22eb

output_dir = '/mnt/d/anima/training_output/light'

dataset = 'dataset-anima.toml'

# training settings
epochs = 1000
# Per-resolution batch sizes
# micro_batch_size_per_gpu = 16
micro_batch_size_per_gpu = [[512, 32], [1024, 32], [1536, 16]]
pipeline_stages = 1
gradient_accumulation_steps = 1
gradient_clipping = 1
warmup_steps = 100

# misc settings
save_every_n_epochs = 1
#save_every_n_steps = 1000
#save_every_n_examples = 4096000
#checkpoint_every_n_epochs = 1
#checkpoint_every_n_minutes = 120
activation_checkpointing = true
#reentrant_activation_checkpointing = true

partition_method = 'parameters'
# partition_method = 'manual'
# partition_split = [10]

save_dtype = 'bfloat16'
caching_batch_size = 1
map_num_proc = 8
steps_per_print = 1
compile = true

[model]
type = 'anima'
transformer_path = '/ComfyUI/models/diffusion_models/anima-preview3-base.safetensors'
vae_path = '/ComfyUI/models/vae/qwen_image_vae.safetensors'
llm_path = '/ComfyUI/models/text_encoders/qwen_3_06b_base.safetensors'
dtype = 'bfloat16'
#cache_text_embeddings = false
llm_adapter_lr = 1e-6
#timestep_sample_method = 'uniform'
#flux_shift = true
#multiscale_loss_weight = 0.5
sigmoid_scale = 1.3

[adapter]
type = 'lora'
rank = 32
dtype = 'bfloat16'

[optimizer]
type = 'adamw_optimi'
lr = 2e-5
betas = [0.9, 0.99]
weight_decay = 0.01
eps = 1e-8</code></pre><pre><code>resolutions = [512, 1024, 1536]

enable_ar_bucket = true
min_ar = 0.5
max_ar = 2.0
num_ar_buckets = 9

[[directory]]
path = '/mnt/d/training_data/images_light_captions'
repeats = 8</code></pre> Dataset format:tags = full tag listfirst_n_tags = kept first 8 tagsnl_caption = natural language captions generated with tag grounding<pre><code># captions.json
{
 "image_1.jpg": [
 {tags},
 {first_n_tags}{nl_caption}",
 {tags}{nl_caption}",
 {nl_caption}{tags}",
}</code></pre>

v3.0 [anima-preview-3]

Trained on <a target="_blank" rel="ugc" href="https://civitai.com/models/2458426/anima-official">Anima Preview 2</a>Assume that any lora trained on the preview version won't work well on the final version. Made some small adjustments to the dataset to include more concepts, like "hygge lighting". First attempt using the same config as v1 on Anima Preview 1Tried some different approaches for Preview 2 training, but they did not produce better results

v2.0 [anima-preview2]

Trained on <a target="_blank" rel="ugc" href="https://huggingface.co/circlestone-labs/Anima">Anima Preview</a> (version 1)Assume that any lora trained on the preview version won't work well on the final version. Trained the mixed tags and natural language dataset.Config:<pre><code># dataset-anima.toml
# Resolution settings.
resolutions = [1024, 1280]

# Aspect ratio bucketing settings
enable_ar_bucket = true
min_ar = 0.5
max_ar = 2.0
num_ar_buckets = 7

[[directory]] # IMAGES
# Path to the directory containing images and their corresponding caption files.
path = '/mnt/d/training_data/images'
num_repeats = 1
resolutions = [1024, 1280]

# Change these paths
output_dir = '/mnt/d/anima/training_output'
dataset = 'dataset-anima.toml'

# training settings
epochs = 50
micro_batch_size_per_gpu = 16
pipeline_stages = 1
gradient_accumulation_steps = 1
gradient_clipping = 1.0
warmup_steps = 100
train_llm_adapter = true

# eval settings
eval_every_n_epochs = 1
eval_before_first_step = true
eval_micro_batch_size_per_gpu = 1
eval_gradient_accumulation_steps = 1

# misc settings
save_every_n_epochs = 1
checkpoint_every_n_minutes = 120
activation_checkpointing = true
partition_method = 'parameters'
save_dtype = 'bfloat16'
caching_batch_size = 1
steps_per_print = 1

[model]
type = 'anima'

transformer_path = '/mnt/c/workspace/models/diffusion_models/anima-preview.safetensors'
vae_path = '/mnt/c/workspace/models/vae/qwen_image_vae.safetensors'
qwen_path = '../qwen0.6/Qwen3-0.6B/'
dtype = 'bfloat16'
timestep_sample_method = 'logit_normal'
sigmoid_scale = 1.0
shift = 3.0

# Caption Processing Options
cache_text_embeddings = false
# NOTE: Requires cache_text_embeddings = false to work!
# For cached embeddings, use cache_shuffle_num in your dataset config instead.
shuffle_tags = true
tag_delimiter = ', '
keep_first_n_tags = 5
shuffle_keep_first_n = 5
tag_dropout_percent = 0.3
protected_tags_file = './protected_tags.txt'

nl_shuffle_sentences = false
nl_keep_first_sentence = true

# 'tags' 'nl' 'mixed'
caption_mode = 'mixed'

debug_caption_processing = false
debug_caption_interval = 1000

[adapter]
type = 'lora'
rank = 64
dtype = 'bfloat16'

# AdamW from the optimi library is a good default since it automatically uses Kahan summation when training bfloat16 weights.
[optimizer]
type = 'adamw_optimi'
lr = 8e-5
betas = [0.9, 0.99]
weight_decay = 0.01
eps = 1e-8</code></pre>

v1.0 [anima-preview]

Light Concepts

motimalu-profile-vid.webp